Section 1: The Basing Relation So-Hyun sees a Chinese Crested dog, and she recalls that hairless dogs that look like that are typically Chinese Crested dogs. At the very same time, her friend Adede points at the dog and says “look at that Chinese Crested dog right there!” So-Hyun believes that the dog is a Chinese Crested. So-Hyun has at least two independent reasons to believe that the dog is a Chinese Crested. One reason is that she heard Adede just say so. And another reason is that, as she recalls, hairless dogs that look like that are typically Chinese Crested dogs. But, even though she has two independent reasons to believe that the dog is a Chinese Crested, it’s at least possible, given my description of the case so far, that only one of those is a reason for which she believes it. This possibility shows that there is a difference between reasons that one has to believe and reasons for which one believes — a distinction in epistemology that is analogous to the distinction that some philosophers of action mark by distinguishing “normative” reasons that one has (reasons that one has to act) from “motivating” reasons (reasons for which one acts). But what does this difference consist in? Donald Davidson tried to explain the difference between reasons that one has to act and reasons for which one acts as a difference consisting in the fact that the latter must be, but the former need not be, reasons that cause one’s action.1 Although this view has been widely accepted, some have objected to the claim that our intentional actions are caused by the reasons for which we do them.2 Rather than get involved in this controversy, let me try to locate Davidson’s insight in a way that does not take on controversial commitments about causation. Davidson’s insight, stated uncontroversially, is this: a reason for whichyou act is always a reason why you act. Or, as some philosophers would put the point, “motivating” reasons are always “explanatory” reasons. This point holds true not just of action, but also of belief: a reason for which you believe is always a reason why you believe. If So-Hyun has two reasons to believe that the dog is a Chinese Crested, and she believes it for only one of those reasons, then the reason for which she believes it must also be a reason why she believes it. Now let me add a stipulation to the case of So-Hyun and Adede: So-Hyun noticed the Chinese Crested dog only because Adede pointed at it and called it a “Chinese Crested” (a phrase that got So-Hyun’s attention, and jogged her recall), but So-Hyun doesn’t at all trust Adede’s judgment in these matters. Adede’s testimony, let me stress, is fully trustworthy, and So-Hyun’s evidence indicates as much, but So-Hyun doesn’t respond rationally to her evidence on this issue. So, although it would be rational for So-Hyun to trust Adede’s testimony, she irrationally fails to do so. In that case, Adede’s testimony is a normative reason to believe that the dog is a Chinese Crested, and, since So-Hyun is aware of her testimony, and has evidence that indicates its trustworthiness, it is also a normative reason that So-Hyun has to believe that the dog is a Chinese Crested. But, in the case I’ve just described, the reason for which So-Hyun believes that the dog is a Chinese Crested is that, as she recalls, hairless dogs that look like that are typically Chinese Crested dogs. The reason for which she believes it is not that Adede pointed it out as such, for So-Hyun doesn’t trust what Adede says. But Adede’s pointing out the Chinese Crested as such is still a reason why So-Hyun believes that the dog is a Chinese Crested, since Adede’s behavior is what explains why So-Hyun notices the dog and recalls what Chinese Cresteds look like in the first place. It follows that, even if all reasons for which we believe are reasons why we believe, still, not all reasons why we believe are reasons for which we believe. So even if you have a reason to believe that is also a reason why you believe, it doesn’t follow that it is a reason for which you believe. Reasons that are both reasons to and reasons why need not yet be reasons for which. And, while all reasons for which are reasons why, not all reasons for which are reasons to. Although we’ve been focused on the case of So-Hyun’s belief, the points that we’ve made generalize, and they generalize beyond beliefs. There are the reasons for which someone raises her hand, the reasons for which she is angry, the reasons for which she intends to drink a toxin, the reasons for which she prefers eating at home to eating out, and the reasons for which she chooses the road less travelled. More generally, there are reasons for which an agent, I will say, is in a “rationally determinable condition” — whether that condition takes the form of a belief, a judgment, an emotion, an intention, a preference, a choice, or an action. The project of this paper is to gain a better understanding of reasons for which — or rather, of the relation that they bear to the rationally determinable conditions for which they are reasons. Epistemologists sometimes use the phrase "the basing relation" to denote the distinctive kind of explanatory relation that there is between a reason and the belief for which it is a reason.I will generalize this usage of “the basing relation” to cover the relation between a reason and the intention, or action, or judgment, or emotion, or choice, or preference for which it is a reason: more generally, it is the relation between a reason and the rationally determinable condition for which it is a reason.Using the phrase in this general way, I can now state the goal of this paper: In this paper, I will give an account of the basing relation. In the next section, I will consider one seemingly plausible account, and argue that it is, at best, incomplete.
Section 2: Basing and Defeat As we’ve told the story, So-Hyun has two reasons to believe that the dog in front of her is a Chinese Crested. One reason is that, as she recalls, dogs that look like that are typically Chinese Cresteds. The other is that Adede said that the dog is a Chinese Crested. But while So-Hyun has both of these reasons to believe it, only one of these is the reason for which So-Hyun believes it: specifically, the reason for which So-Hyun believes it is the first, but not the second, of the two reasons just enumerated. In virtue of what is one, but not the other, a reason for which So-Hyun believes it? I’d now like to articulate one plausible proposal. Consider what happens if So-Hyun gets evidence that, contrary to her recollection, Chinese Cresteds do not typically look like the dog in front of her. This evidence will defeat the doxastic justification of So-Hyun’s belief that the dog in front of her is a Chinese Crested. One indication of this defeat is that it would typically be rational for So-Hyun to respond to such evidence by reducing her confidence, or perhaps even suspending her belief, that the dog in front of her is a Chinese Crested. But now consider what happens if So-Hyun gets evidence that Adede did not say that the dog was a Chinese Crested. This evidence will not defeat the doxastic justification of So-Hyun’s belief that the dog in front of her is a Chinese Crested. One indication of this lack of defeat is that it would not be rational for So-Hyun to respond to such evidence by reducing her confidence that the dog in front of her is a Chinese Crested. According to the present proposal, it is her recollection of the appearance of Chinese Cresteds, but not Adede’s testimony, that is the reason for which So-Hyun holds her belief, and this is because the doxastic justification of So-Hyun’s belief that the dog is a Chinese Crested can be defeated by defeating her justification for believing the former reason, but cannot be defeated by defeating her justification for believing the latter reason. Just as beliefs can be doxastically justified, so too can intentions, actions, choices, preferences, and emotions be justified. More generally, rationally determinable conditions (henceforth, RDC’s) can be justified by virtue of being based in the right way on justifying reasons. I take doxastic justification therefore to be just one species of a broad genus, and I will use the phrase “RDC justification” to denote this genus. Just as doxastic justification can be defeated, RDC justification more generally can be defeated. And just as the defeat of doxastic justification is typically indicated by its being rational for the agent to suspend belief, so too the defeat of RDC justification is indicated by its being rational for the agent to suspend her RDC. “RDC justification” then, as I use the phrase, is not the justification that an agent has for being in a particular RDC. It is rather the justifiedness of the RDC itself – the RDC that the agent is in. Three remarks about what’s involved in RDC justification:
At least sometimes, when an agent A is in a particular RDC, and that RDC enjoys RDC justification, the latter is true partly (but only partly) because A has a normative reason to be in that RDC. For instance, my intention to wipe the windshield is justified because I have a reason to intend to wipe the windshield; my belief that the building is on fire is justified because I have a reason to believe that the building is on fire, etc. In these cases, RDC justification requires at least the agent’s having a normative reason to be in that RDC.
But under no circumstances is having a normative reason to be in a particular RDC sufficient for that RDC enjoying RDC justification. Having a reason to believe that the building is on fire is not sufficient for my belief that the building is on fire to be justified; having a reason to want to wipe the windshield is not sufficient for my desire to wipe the windshield to be justified. If a normative reason is involved in a particular case of RDC justification, it must be involved as a certain kind of basis of the RDC. The normative reason is never, by itself, sufficient for RDC justification.
Even if an agent has a normative reason to be in a particular RDC, and that normative reason is the basis for the RDC itself, that still is not sufficient for RDC justification. Consider the following example from Turri 2010:
“Mr. Ponens and Mr. F.A. Lacy each knows the following things: (P5) The Spurs will win if they play the Pistons.
(P6) The Spurs will play the Pistons. This is a paradigm case of propositional justification. is propositionally justified for each man because he knows (P5) and (P6). …From these two premises, and only these premises, each man draws the conclusion: (P7) Therefore, the Spurs will win. …But the devil is in the details. Ponens applies modus ponens to reach the conclusion. Lacy, however, applies a different inference rule, which we may call modus profusus: for any p, q, and r: (p^q) –> r. Lacy’s belief that the Spurs will win is definitely not doxastically justified; following that rule could never lead to a justified belief.”3 Such examples illustrate that having a normative reason – even a completely decisive, ultima facie normative reason, to be in a particular RDC, and basing one’s RDC on that normative reason, does not suffice for one’s RDC to be justified. As Turri persuasively argues, achieving RDC justification on the basis of a normative reason to be in that RDC, the basing relation must be of the appropriate kind (e.g., proceeding through modus ponens, rather than modus profusus). But what distinguishes basing relations that are of the appropriate kind from those that are not? Is it that basing relations are of the appropriate kind if they proceed in accordance with valid rules of inference (as in the modus ponens vs. modus profusus case above)? That cannot be right, since it would mean that beliefs that result from inductive or abductive inference could not be doxastically justified, and this is plainly false. Is it then that basing relations are of the appropriate kind if they proceed in accordance with good rules of inference? We cannot evaluate this proposal until we say more about what it is for a rule of inference to be “good”, and it’s not yet clear what more we can say (though the view I develop below will eventually put us into a position to say more). We are not yet in a position to answer the question of what distinguishes basing relations of the kind that can generate RDC justification from basing relations that are not of this kind. This is a question that we will eventually answer, by appeal to our eventual account of the basing relation, but we are not yet in a position to answer it. So for now, we’ll adopt the terminological convention of referring to basing relations of the former kind as “proper” and basing relations of the latter kind as “improper”. These are just labels for now, to be understood for now only by appeal to examples like the one from Turri; eventually, we will have an account of the basing relation that will explain these labels. Now that we’ve said something to identify the concept of RDC justification, I can use that concept to propose the following plausible, but ultimately unsatisfactory, account of the basing relation: R is a reason for which A is in rationally determinable condition C = C can have its RDC justification defeated by defeating A’s justification for accepting R. Although this account of the basing relation appeals to normative terms (like justification anddefeat) in order to explain basing, this does not strike me as problem with the account: there’s no reason to think that we can or should try to explain basing in non-normative terms. Furthermore, this account seems to make at least many, if not all, of the right predictions concerning what stands in the basing relation to what. It is possible, for all I say in this paper, that there are cases in which we might wish to say that a creature believes, or intends, or acts, for particular reasons, but for which the account above makes the wrong predictions: but if there are such cases, I’m inclined to think that they show only that our ordinary use of the phrase “reasons for which” is poorly regimented. In particular, the account above will be subject to a particular complaint that will also be predictably leveled against the account that I will eventually propose below, viz., that it makes the basing relation one that can obtain only in creatures who occupy rationally determinable conditions, and whose rationally determinable conditions can (as all RDC’s can) be assessed as more or less rational. This, it will be said, is to overintellectualize the basing relation. Although I know that this charge will be leveled, I believe that it misunderstands our explanandum. Are there reasons for which the fly moves towards the light, or are there only reasons why it does so? I think the latter, but if I am wrong about this, then that shows only that the phrase “reasons for which” fails to specify my explanandum as precisely as I would like, then I will specify it more precisely now: the basing relation that I’m interested in understanding is a relation that bears on the justification of a RDC.
So, putting aside the overintellectualization objection as irrelevant to our present explanandum, the account I’ve stated above seems, at least initially, quite plausible. Are we done? No. This account, even if extensionally correct, suffers from two shortcomings. The first is that the account explains things the wrong way round. The fact that A’s C’ing can have its RDC justification defeated by defeating A’s justification for accepting R seems to be explained by the fact that R is a reason for which A C’s, but the account says that the former is what explains the latter. The second problem, which is related to the first, is that the account fails to explain a puzzling and important phenomenon concerning the defeat of RDC justification. In the remainder of this section, I will state this phenomenon, and then say why the account proposed above fails to explain it. As we’ve told the story about So-Hyun, the reason for which she believes that the dog is a Chinese Crested is that, as she recalls, hairless dogs that look like this are typically Chinese Crested. Here’s a diagram:
Now consider the variety of ways in which So-Hyun could fit the description that I’ve given, and nonetheless be unjustified in believing that this dog is a Chinese Crested. This could happen if So-Hyun has an opposing, or overriding, defeater to which she is insufficiently sensitive: for instance, she could have, and ignore, independent evidence that the dog in front of her is not a Chinese Crested. Such a defeater directly attacks the rightmost element in the picture above. Or she could have an undercutting defeater to which she is insufficiently sensitive: for instance, she could have, and ignore, independent evidence that her recall is very poor when it comes to information about the appearance of dog breeds. Such a defeater directly attacks the leftmost element in the picture above. Or, finally, even if she is fully justified in believing that dogs that look like that typically are Chinese Cresteds, she could still have, and ignore, independent evidence that the particular dog in front of her is very atypical of its breed. The first kind of defeater can oppose her justification for believing that the dog is a Chinese Crested, even though it does nothing to defeat her justification for believing that hairless dogs that look like this are typically Chinese Crested. The second kind of defeater can undercut her justification for thinking that the dog is a Chinese Crested by virtue of defeating her justification for believing that hairless dogs that look like this are typically Chinese Crested. But the third kind of defeater has a different effect from either of the others: it defeats her justification for thinking that this dog is a Chinese Crested, but it does not oppose her justification for thinking this, nor does it defeat her justification for believing that dogs that look like this are typically Chinese Cresteds. How does this work? How can her justification be defeated without being opposed, and without defeating her acceptance of the reasons that supply that justification? This kind of defeater would need directly to attack the middle element in the picture above, rather than the rightmost or leftmost element. But how should we understand the middle element, so as to make sense of this possibility of defeat?4 It may seem that there is an obvious answer to our questions about how this third kind of justification defeat or augmenting works — namely, that we have so far identified So-Hyun’s reason for her belief much too narrowly, as consisting merely in the fact (or proposition, or state of apparent recollection) that dogs that look that way are typically Chinese Cresteds. But this latter — it might be thought — is just the tip of a whole iceberg of reasons that So-Hyun has for her belief. Once we expose the rest of the iceberg, it will become obvious that there’s nothing unusual about the kind of justification defeat or augmenting that we’ve considered: it is nothing other than the kind of justification defeat or augmenting that we get when we add a new piece of evidence to an agent’s total body of evidence, and thereby affect the agent’s degree of justification for some RDC that she makes on the basis of this total body of evidence. This response is too quick. Suppose we specify the whole of So-Hyun’s reason to believe that the dog is a Chinese Crested — however extensive that whole body of reasons is. Indeed, let it be her total body of evidence, and let her belief be doxastically justified, by virtue of being based in the right way on this total body of evidence. Now suppose that So-Hyun gains one additional piece of evidence, and it leaves her justification for accepting the truth of every proposition in that whole body of evidence completely unaffected, but it does affect her justification for thinking that that the whole body of evidence supports her belief that the dog in question is a Chinese Crested. Perhaps an eminent mind-reading statistician assures her that, much as it may seem to her as if the rest of her total evidence supports the proposition that the dog in question is a Chinese Crested, in fact it does not do so. This new piece of evidence need not affect the justification that So-Hyun has for accepting any of the rest of her evidence; nonetheless, it defeats the doxastic justification of her belief that the dog in question is a Chinese Crested.5 So, no matter how extensive So-Hyun’s reasons to believe that the dog is a Chinese Crested, the doxastic justification of her belief can be defeated (or augmented) without affecting her justification for accepting those reasons. It is typical for epistemologists to distinguish opposing defeat from undercutting defeat. But what this discussion has shown is that there are actually three kinds of defeat. Opposing defeat: C’s RDC justification is defeated in virtue of reasons to not C, and independently of any effect on the justification of her normative reasons to C Undercutting defeat: C’s RDC justification is defeated in virtue of defeating her justification for accepting her normative reasons to C Side defeat: C’s RDC justification is defeated in virtue of reasons to doubt that it is properly based on the agent’s normative reasons to C, and independently of any effect on the justification of those normative reasons, and independently of any other reason to not C.6 Pictorially:
The question I want to raise now is: why is there such a thing as side defeat? In other words, why does reducing an agent’s justification for thinking that her reasons provide justification for her belief work to defeat the justification for her belief? Notice, by the way, that the standard distinction between higher-order defeaters and first-order defeaters cuts across this three-fold distinction between opposing, undercutting, and side defeat. There could be first-order opposing defeat, higher-order opposing defeat, first-order undercutting defeat, higher-order undercutting defeat, and so on. The account of the basing relation proposed at the beginning of this section — the account that explains basing in terms of undercutting defeat — does nothing to help us understand how side defeat works. But, just as basing is related to undercutting defeat, so too it is related to side defeat: if R is a reason for which A C’s, then the justification of A’s C’ing can be defeated not simply by defeating R, but also by gaining reasons to doubt the connection between R and C’s RDC justification. If we’re going to understand basing in terms of its relation to defeat, then we need to understand why it is related in the way that it is not merely to undercutting defeat, but also to side defeat. Once we understand how basing is related to side defeat, we will then also be in a position to explain the distinction introduced above, in connection with Turri’s example, between proper basing (that provides for RDC justification) and improper basing (that doesn’t).